Changelog.md

CARE 2.3

• Faster merge step during construction of output file
• Use CCCL 2.2.0
• Update program options for GPU version
• GPUs are now specified via CUDA_VISIBLE_DEVICES environment variable
• Massive refactoring to improve both single-GPU performance and greatly improve multi-GPU performance
• Bugfixes:
  • Fix potential race in gpu msa
  • Fix memory consumption of gpu hash tables

CARE 2.2

• CPU / GPU performance improvements
• Add option to specify data layout for hash tables and reads when using multiple gpus
• Add option to specify detailed thread configuration for gpu version
• Bugfixes
• Refactoring

CARE 2.1

• .gz output
• Avoid some non-deterministic floating-point operations in gpu version
• Faster hash table construction in gpu version
• Improve multi-gpu performance
• Update dependencies
• Bugfixes
• Refactoring

CARE 2.0

• Add a dedicated code path for paired-end inputs
• Add a Random Forest-based correction mode
• Add an option to compress quality scores
• GPU version can use GPU hashtables
• Big code refactoring
• Performance and memory improvements